Technologies for accepting speech uttered from a user as input speech and performing speech recognition on the input speech to recognize strings from input speech have been recently developed. However, errors may be likely to occur in speech recognition. In this regard, there has been developed a technique of generating a plurality of processing units except some strings from strings recognized from the input speech, presenting them to a user, and regarding a processing unit selected by a user from a plurality of processing units as a target to be processed (e.g., refer to Patent Literature 1).